Strategic Default in joint liability groups:
Evidence from a natural experiment in India∗
Xavier Giné
World Bank
Karuna Krishnaswamy
CGAP
Alejandro Ponce
World Justice Project
November 2011
PRELIMINARY AND INCOMPLETE
Abstract
Despite the high repayment rates claimed by microcredit programs around the world,
some groups of borrowers eventually default and are subsequently disbanded. Exposure
to common shocks and strategic default are reasons for the deterioration in group
repayment but identification of the precise mechanism is difficult. In this paper we
exploit an announcement issued by the Anjuman Committee of a town in southern India
banning all Muslims from repaying their microfinance loans. Using administrative data
we find that borrowers in Muslim-dominated groups have higher default rates after the
announcement compared to the same borrowers with loans in Hindu-dominated groups.
We conclude that strict adherence to joint liability rules may have triggered strategic
default that might have been avoided if lenders had allowed more flexibility in
repayment.
JEL: C81, G21, O12, O16
∗
Giné: xgine@worldbank.org. Krishnaswamy: karuna.krishnaswamy@gmail.com. Ponce:
aponcer@gmail.com. We thank Jordi de la Torre for superb research assistance. We are grateful to David
McKenzie, Maitreesh Ghatak, Dean Karlan, Bilal Zia and conference participants at BREAD, GUC in
Cairo and the World Bank. The views expressed in this paper are those of the authors and should not be
attributed to the World Bank, its executive directors, or the countries they represent.
1
1. Introduction
In the last 25 years, microcredit has arguably become one of the most popular
tools among governments, NGOs and multilateral institutions to make credit available
to low-income households who lack access to formal credit markets. In 2009, the
Microcredit Summit estimated that there were more than 3,500 microfinance institutions
around the world with 150 million clients (Daley-Harris 2009). One unique feature of
micro loans, pioneered by Nobel laureate Muhammed Yunnus and the Grameen Bank,
is “group lending”, defined as the provision of loans to individual clients who are part
of a small group (typically comprised of 5 to 20 members) and meet regularly to repay
their loans.
Group lending has been credited for the high repayment rates claimed by
programs around the world. Group lending loans contracts often include a joint liability
clause but this is not always the case. 1 Joint liability conditions future loans to group
members to the repayment of the group as a whole by requiring that all members in a
group be responsible for the loans of each other. If one member cannot meet his or her
repayment obligation, other members must bear the repayment of the defaulter if they
want to continue borrowing from the lender.
Theorists have been particularly interested in the repayment incentives induced
by joint liability. 2 The main advantage is that it solves informational asymmetries by
shifting the burden from the lender to the clients resulting in lower transaction costs for
the institution (Ghatak and Guinnane, 1999), thus providing a way around the common
problems of adverse selection (screening and sorting) and moral hazard (ex-ante and expost). 3 In theory, joint liability contracts can lead to higher repayment because
borrowers have better information about each other’s types, can better monitor each
other’s investment, and may be able to impose social sanctions at low cost. Joint
1
Grameen Bank has recently introduced a new product, Grameen II that removes joint liability (Dowla
and Barua, 2006). The Association for Social Advancement (ASA), one of Grameen’s main competitors
in Bangladesh has also abandoned joint liability while retaining the practice of meeting with clients in
public groups. See Giné and Karlan (2011) for more details on the recent trends.
2
The theory on joint liability builds on the contract theory literature from the early 1990s that studies
when a principal should contract with a group of agents rather than individually with each agent. See for
example Holmstrom and Milgrom (1990), Varian (1990) and Arnott and Stiglitz (1991).
3
See for example Ghatak (1999; 2000), N’Guessan and Laffont (2000), Sadoulet (2000) and Armendariz
de Aghion and Gollier (2000) for models of adverse selection, Stiglitz (1990) and Laffont and Rey (2000)
for models of ex-ante moral hazard and Besley and Coate (1995) for a model of ex-post moral hazard. For
reviews see Morduch, 1999 and Armendariz de Aghion and Morduch, 2005.
2
liability, however, has its own pitfalls. 4 Attendance in group meetings and monitoring
group members can be costly, especially in areas with low population density. 5 More
importantly, Besley and Coate (1995) argue that the whole group may default, even
when some members would have repaid under individual liability. This happens when
the number of defaulting clients in the group is so large that the remaining members
cannot afford the repayment of defaulters, along with their own repayment. In this
situation, borrowers that could repay their loans have little incentive to do so because
access to future loans will be denied. As a result, they will strategically decide to
default.
Because joint liability is embedded in group lending schemes, we have little
evidence on its relative importance for repayment vis à vis other features, such as the
regular meetings where repayment is public. In one of the two field experiments
reported in Giné and Karlan (2011), the lender removed joint liability from pre-existing
groups while maintaining the weekly meetings. After three years, they find no increase
in short-run or long-run default perhaps because the group meetings were still
encouraging some monitoring and enforcement due to reputation or shame. Indeed, Rai
and Sjöström (2000, 2010) and Feigenberg et al., (2010) provide evidence that group
meetings may be helpful beyond joint liability. 6
More generally, while some groups do default and disband, we have little
evidence about the actual reasons for the deterioration in repayment. Groups may
default for various reasons, most notably because they are exposed to common shocks
and for the strategic reasons argued in Besley and Coate (1995). Understanding the
actual mechanism is important for the design of microcredit products that minimize the
occurrence of group default.
In this paper we provide evidence of strategic default using a natural experiment
in southern India. In particular, we study whether members of a joint liability group are
more likely to default on their loans when the proportion of defaulting members in a
4
On anecdotal evidence on the limits to joint liability, see Matin (1997), Woolcock (1999) Montgomery
(1996) and Rahman (1999).
5
Park and Ren (2001) find that 7% of microfinance clients in some programs in China have to travel
more than one hour to attend the group meeting.
6
First, public repayments at group meetings can strengthen the strategic use of social
stigma in aid of the programs' bottom lines, even without a formal joint liability contract. Second, the
group is often a useful resource through which staff can directly elicit information about errant borrowers
and create pressure as needed (Rai and Sjöström, 2000). Finally, Feigenberg et al., (2010) find that an
increase in the frequency of meetings leads to increased risk sharing and social interaction outside of
meetings.
3
group increases. Despite the simplicity of this hypothesis, in the absence of an
exogenous source of variation it becomes difficult to identify this mechanism, since
repayment rates are the result of selection, incentive effects and correlated observed and
unobserved shocks. Our identification strategy overcomes these concerns by exploiting
two facts of the data. The first is an unexpected event that increased the default rate
among Muslims but not Hindus. On Jan 28th, 2009, the Anjuman Committee of the
town of Kolar in the state of Karnataka, India, issued a fatwa banning all Muslims from
repaying their microfinance loans claiming that charging interest was “haram”
(forbidden). The ban led to immediate non-repayment by Muslims clients. As a result,
borrowers in Muslim-dominated groups faced, after the ban, a greater repayment burden
compared to borrowers in Hindu-dominated groups. The second fact is that in our
setting many borrowers had loans from several groups, which differed in the density of
Muslims. Because borrowers from Muslim-dominated groups may be inherently
different those in Hindu-dominated groups, we also focus on borrowers with multiple
loans. Identification in this case comes from the variation in the behavior of the same
individual across multiple groups of differing density of Muslims.
Using matched administrative records of 7 of the largest MFIs operating in the
state of Karnataka, India, we find that the after the ban, the likelihood of default at
maturity for Hindus and Muslims in mixed religion groups increases with the
percentage of Muslims in the group. In particular, a one percentage point increase in the
share of Muslims contributes to a 0.5 percentage point increase in the likelihood of
default after the ban. This result is robust to alternative measures of default, such as the
balance outstanding at maturity or the percentage of the loan due at maturity.
Using survey data we can rule out several other explanations for the observed
mass defaults. First, one could argue that borrowers and specially Muslims wanted to
pay but that they were physically unable to do so. Interviews with credit officers suggest
that the meetings took place and that they were always willing to meet to collect
repayment. Second, it could be also argued that the fatwa lowered the non-pecuniary
cost of defaulting by turning it into a less shameful act. Similarly, the fatwa may have
provided individuals with an idea about the diminished consequences of defaulting. Put
differently, individuals may have realized that default is more acceptable or has little
negative consequences. But if this were the case, then borrowers with multiple loans
would default in all groups, and not only those with high density of Muslims, which is
what we find. Survey data provides one last piece of evidence in support of the strategic
4
motive for default. Individuals that had missed at least a payment were asked for the
reason and most of the respondents, irrespective of the religion mentioned that they had
the money to repay the loan but chose not to.
These findings contribute to the broader literature on peer effects and in
particular to the literature that studies the causal impact of peer’s behavior on own
repayment behavior. In this sense, this paper is closest to Breza (2011) who study an
earlier episode of mass defaults in the Krishna district of Andhra Pradesh, India.
The remainder of this paper is organized as follows. Section 2 provides the
context of the natural experiment. Section 3 describes the data. Section 4 describes the
identification strategy for our analysis and its results. Section 5 concludes.
2. Context
2.1. Institutional Setting
The state of Karnataka in South India is home to a competitive microfinance
industry that implements the Grameen-style model. At the time of the fatwa, there were
27 registered and many unregistered microfinance institutions operating in rural and
urban areas (EADA, 2010).
Institutions typically open a branch office in an urban or semi-urban location and
serve approximately 3,000 to 5,000 clients. Clients are almost entirely women and
reside in town and in adjoining villages and smaller towns. Clients of a branch are
organized into groups and centers. Loan officers form and train groups of between 5 and
15 borrowers. Members in a group typically reside in the same neighbourhood or street
and know each other so that they can verify appropriate loan utilization and monitor
effort into the member’s chosen project. Members of a group meet weekly at a center, at
a public location in the colony. The center meeting is managed by the loan officer,
called a center manager. A center has approximately 40 members formed by groups
residing in the same colony. The center manager disburses loans and collects weekly
instalments at the center meeting. Attendance and repayment are strictly enforced.
Typically none of the members is allowed to leave the meeting until all the collections
have been made.
5
Three types of loan products are typically offered: an income generation loan
used for investment, a supplementary or top-up loan four to six months after the
disbursement of the income generation loan and an emergency loan given on the spot
and of smaller size for consumption purposes. Loans range from Rs. 500 to Rs. 30,000
depending on the loan cycle, and repayment starts one week after the date of loan
disbursal and is spread over 50 weekly instalments. Interest rates vary from 12.5% to
30% and are charged on a declining balance. All loans are guaranteed by joint liability
and none of the microfinance institutions can collect savings that could be used as
collateral. Loans are transacted and recorded on an individual basis, but an outstanding
default by any group members renders the entire group ineligible for future loans.
2.2. A natural experiment
By 2008, the explosive growth of MFIs in southern Karnataka started to overheat. In the
quest to meet their growth targets, loan officers often disbursed loans to clients already
indebted to other organizations. When conditions in the silk reeling industry
deteriorated (partly on account of the global crisis), the strain of repaying large sums of
money on a weekly basis became excessive for some households. 7
The situation took a religious turn when friends and relatives of affected women
complained to the local religious establishment about the “trouble” caused by MFIs.
The local Anjuman Committee –irked by months of complaints about women
“neglecting” family duties, families led into crippling debt by a culture of “easy
money”, and oath taking at meetings invoking Hindu goddesses– intervened in the only
way it knew: interaction between MFIs and Muslims in the town was prohibited by
religious edict. On Jan 28th, 2009, triggered by the attempted suicide of a prominent
member of a member community in Kolar whose wife had become over-indebted, the
Anjuman Committee of Kolar issued a statement banning all Muslims from repaying
their MFI loans claiming that charging interest was “haram”. The ban lead to immediate
non-repayment by Muslims clients. The situation eventually led to a complete
breakdown of interaction between MFIs and their clients in Kolar. Repayment issues
extended shortly to the towns of Ramanagaram and Mysore.
7
Discussion with clients in the four main affected towns shows that, in many cases, they were running
from one MFI meeting to another spending 1-2 hours on a daily basis.
6
The resulting delinquency crisis amongst Muslim clients spread to other towns as a
variety of vested interests came into play. In nearby Sidlaghatta, the religious ban of
Kolar was sufficient to provide a breather from the daily round of financial juggling that
went into repaying multiple microfinance loans without an adequate income from their
household reeling enterprises. In Ramanagaram, reeling factory owners faced with a
labor shortage on account of the increasing independence of women with microfinance
loans, appear to have influenced the local Anjuman to follow the Kolar example and
ban the interaction of Muslims with MFIs. In Mysore, an unrelated communal clash
resulted in business losses that enabled a local political organization to pose as a savior
of the community by urging microfinance borrowers not to repay and by raising the
possibility of a loan waiver.
The situation was compounded by a zero delinquency policy, which did not give MFI
staff the flexibility to negotiate or reschedule payments when clients got into trouble.
MFI staff required invariably to collect installments on time, which put pressure on
clients to repay. Equally importantly, the mass default having occurred, the group
liability mechanism had unintended consequences. Even those members of a group
willing to pay did not see any point in doing so as they were branded as defaulters along
with other members of their group, with no prospect of persuading MFI staff to consider
their case separately. Members of mixed groups not otherwise defaulting were
particularly affected by this phenomenon.
The fatwa provides us with exogenous variation, as it is arguably unrelated to the
unobservable factors that drive default. It caused an unanticipated default of the
majority Muslims members without affecting Hindu clients
3. Data
We use two sources of data. First, the administrative records from seven out of
the eight largest microfinance institutions operating in the towns of Kolar and
Ramanagara in the state of Karnataka, India. 8 Four institutions provided all the loans
taken since 2007 until December 2009 while three provided data since 2008 until
December 2009. Second, we use survey data collected in Kolar and Ramanagara from a
stratified sample of about 800 households. The data include information on borrowers’
8
We obtained data from Grameen Koota, SKS, Ujjivan, RORES, FFSL, Asmitha and Spandana. We do
not have data from Bharati Swamukthi Samsthe (BSS).
7
behavior and the lenders’ practices. The stratified sample was drawn from the
administrative data set.
Since many customers borrow from multiple institutions, we matched the
administrative records across institutions by name and husband name using a phonetic
algorithm given that there is no credit bureau and the spelling of names may be different
in different databases. Although ID checks are part of the enrolment process, the
algorithm is not perfect because clients sometimes use nicknames and other short forms
that will not be matched. 9
An observation is the master administrative data is a loan with maturity between
August 2008 and December 2009 granted by one of the institutions to an individual.
The information available includes (i) branch, group and center identifiers of each
client; (ii) loan characteristics such as amount, interest rate, frequency of repayments,
disbursement date, and date of maturity; (iii) whether the loan fully repaid at maturity
and if not the amount remaining to be repaid and; (iv) basic socio-economic information
about the borrower such as name, husband name, age, colony, town or village, religion
and caste. The information collected varies slightly from one institution to another. In
our analysis, we use only variables reported by all institutions. It is important to note
that defaults are recorded on an individual loan basis although joint liability is enforced
at the center level. A closer inspection of the data provides many examples of some
members of the group defaulting while the rest do not.
Our sample contains information about 47,794 clients. In our analysis, however,
we use loans disbursed only before Anjuman fatwa. This leaves us with 54,435 loans
among 33,862 borrowers in 2,531 groups.
Table 1 provides the main descriptive statistics for our sample of loans and
individuals. We report different three subgroups of Hindus and Muslims in addition to
the full sample in columns 1-3. Columns 4-6 report only borrowers (both Hindu and
Muslims) that borrow at least once from a group with religious diversity (mixed center).
Columns 7-9 report the subset of borrowers with multiple loans and columns 10-12
9
We evaluated the effectiveness of our algorithm by comparing the results obtained by a manual match
on 1,000 random individuals. In approximately 95 percent of the cases, the algorithm provided a correct
match.
8
report the intersection of the previous sets of borrowers, namely, the subset of clients
that borrow from mixed centers and that have multiple outstanding loans at some point
in time between August 2008 and December 2009.
Table 1 shows that a little more than half of all clients are Hindus (53 percent)
while the remaining 47 percent are Muslims. Approximately 42 percent of all loans are
disbursed to clients in mixed religion groups and a bit less than a quarter of all loans are
issued to individuals with multiple loans. These borrowers have more than 3
outstanding loans on average during the study period. Finally, around 13 percent of
loans are issued to individuals with multiple loans in mixed centers. The average loan
size in the sample is 8,500 Rs (around USD 177 at the time of the ban) and the average
duration is approximately 330 days. These figures vary little from one sample to
another.
4. Empirical Strategy and Results
4.1 Validity checks
We first verify that our differences in differences approach is valid. First, we
rule out possible alternative explanations for the surge in defaults among Muslims
observed after the fatwa. One key concern is that Muslims face a larger debt burden
after the fatwa compared to Hindus. The higher burden could be the result of more
loans, higher loan sizes, higher interest rates, lower number of installments or a
combination of any of these reasons.
Figures 1, 2, 3, and 4 show that the number of loans, the average loan size, the
interest rate, and the number of installments, respectively, do not vary significantly
before and after the announcement, suggesting that indeed, the Anjuman Committee ban
is responsible for the surge in loan defaults.
Second, we need to ensure that joint liability at the center level, and in particular
the threat of future credit denial, is actually enforced. To this end we check whether
there is a drop in loan disbursement after the fatwa, following the increase in defaults.
Figure 5 shows that this is indeed the case. Similarly, Figure 6 shows that there is a
steep decline in the size of the loans disbursed after the fatwa.
4.2 Regression specification
9
In our analysis we use the following regression differences in differences
specification with standard errors clustered at the center level. For loan l of individual i
in center c at time t:
Ylict = aPt + bSc + d Pt x Sc + Dic + εlict
where Ylict is the outcome of interest, Pt is an indicator for the Post announcement
period, Sc is the share of Muslims in center and Dic are individual fixed effects.
We focus on three repayment outcome variables, namely, a dummy for default at
maturity, the balance outstanding at maturity, the percentage of loan due at maturity.
We also include the loan amount to verify that there were no differences in
disbursement.
One possible concern with the repayment variables is that after maturity the loan
could have been repaid. Although we do not have direct data, credit officers have told
us, that very few loans were repaid after maturity.
4.3 Main results
We report the results separately for Hindu and Muslim borrowers. Within each
group of borrowers, we use four different subsamples to better distinguish the extent of
strategic default among our population. Table 2 shows the results of our main
regressions with default at maturity as the dependent variable, for the subsample of
Hindus. The first column includes all loans held by Hindus, column 2 only those loans
from Hindus in mixed groups (hence excluding loans from Hindus in groups with no
Muslims). The sample in column 3 is that of loans given to Hindus with more than one
loan and, finally, column 4 provides the cleanest evidence of strategic default by using
only loans extended to Hindus who have loans in more than one (mixed) center. As
mentioned before, this last approach offers a good way to overcome the consequences
of likely differences among groups beyond the variation in the density of Muslims. For
all subsamples, loans maturing after the fatwa (recall that the regressions in this table
use only loans extended to Hindus) experience a significantly higher default rates than
those maturing before. The size of this effect ranges from 10 to more than 20 percentage
points. What’s more, the restricted sample of those Hindus who hold multiple loans in
more than one group (with mixed membership) yields a still higher point estimate of
10
this parameter. Those are the observations that capture in a cleaner way the effect we
are trying to establish. Still more important for our purposes, our regressions show a
strong, positive and significant effect for the interaction of the variable post (loan
maturing after the fatwa) and the share of Muslims in the group. This can be understood
as evidence for strategic default, since even after the fatwa those groups with a higher
share of Muslims are also those where Hindus default in greater proportion. Our results
indicate that a one percentage point increase in the share of Muslims in a given group
contributes, after the fatwa, to a 0.5 percentage point increase in the likelihood of
default for a loan to a Hindu borrower in that group. The results hold for Muslim
borrowers as well (Table 3): the larger the share of Muslims in a given group, ceteris
paribus, the larger the default rate after the fatwa. Shifting to the balance outstanding at
maturity as our dependent variable (Table 4 and Table 5), the story remains exactly the
same: for Hindu borrowers, the higher the share of Muslims in a group, the higher this
balance is after the fatwa. The effect holds, albeit at a slightly less strong pace, for the
class of Hindu borrowers with multiple loans outstanding in groups with mixed
membership. For Muslim borrowers, the effect is also present, but ceased to be
statistically significant for the categories of borrowers with multiple loans and those
with multiple loans in mixed groups.
4.4 Alternative specification
In the spirit of Besley and Coate (1995) we now run a specification to estimate the
critical mass of defaulting members in a group that triggers strategic default among
members that could otherwise repay. An estimation of this parameter is valuable for the
design of microfinance schemes in general.
We use an alternative specification that relies on the use of categorical variables for the
density level of Muslims in the different groups. We define four Muslim-density
dummy variables: Dens1-Dens4. Dens 1 takes value 1 if the group has no Muslim
presence. Dens2 identifies those groups with a Muslim density below 6.25%, Dens2
between 6.25 and 25% and Dens4 above 25%. As before, the econometric specification
is the following, replacing the variable standing for the density of Muslims with the new
density dummies:
11
Ylict = aPt + bdens2 c + cdens3 c + ddens4 c + e Pt x dens2c + f Pt x dens3 c + g Pt x
dens4c +Dic + εlict
Also as before, we split the sample by loans belonging to Muslims and to Hindus, but
we only use the subsample of loans from borrowers with multiple loans in mixed
groups. We use as dependent variables whether the loan was in default at maturity, the
balance outstanding at maturity, the percentage of the loan due at maturity, and the loan
principal amount.
We report the results in Table 6. Our findings point to a strong and significant effect on
loan repayment among Hindus for the two categories of groups with the highest density
of Muslims. In particular, a loan from a Hindu belonging to a group in the third
category (Muslim density between 6.25% and 25%), with a maturity date after the
fatwa, was 17.7 percentage points more likely to be in default at maturity than the
average loan in the same situation in other groups. Equally, those in the fourth category
(Muslim density above 25%) were 28.5 points more likely to be in default. Similarly
significant effects are found for balance outstanding at maturity, with loans of Hindus in
groups of high Muslim-density showing significantly higher balances than the rest.
4.5 Summary of results
To sum up, our analysis shows convincing evidence that microfinance customers in
joint liability schemes do in fact engage in some measure of strategic default, hence
providing some corroboration for the Besley-Coate hypothesis. Both Hindus and
Muslims in Muslim-dominated groups show higher default rates. For all the outcomes
studied, the evidence suggests that borrowers are more likely to default on their loans
when the fraction of defaulters in their group rises.
Our exercise identified large differences in default rates for borrowers in Muslimdominated centers relative to borrowers in Hindu-dominated groups. But the question
remains, what are the precise mechanisms through which this effect takes place?
To complement our analysis of administrative records of loan performance with a
better picture of the fundamental reasons behind the phenomenon, we draw from a
survey (more details of survey) conducted in the area around the same time. In that,
borrowers were asked the reason of their missing payments. 95 per cent of Muslims and
12
89 per cent of Hindus answered that they actually had the money to repay the loan, but
chose not to. This is precisely the strategic behavior we are concerned about. Strategic
default seems to be the leading explanation for the surge in loan default, as opposed to
shocks or institutional constraints.
6. Conclusion (to be completed)
The benefits of joint liability schemes have been well documented for long (references).
Joint liability, however, is not a panacea and it does not come without its share of
problems. As suggested by Besley and Coate, and reinforced by our analysis, strict
adherence to joint liability rules can have unintended consequences that should be taken
into consideration when choosing among alternative schemes to help the poor overcome
their lack of access to credit. Among them, the possibility of strategic default seems
likely to merit close attention.
Evidence of strategic default has already triggered some changes for MFIs. According
to Dowla and Barua (2006), suche evidence in Grameen group liability groups triggered
conversion to invididual liability under Grameen II.
References
Ahlin, C. and R. Townsend (2007). "Using Repayment Data to Test Across Models of
Joint Liability Lending." Economic Journal. Forthcoming.
Armendariz de Aghion, B. and C. Gollier (2000). "Peer Group Formation in an Adverse
Selection Model." The Economic Journal 110(465): 632-643.
Armendariz de Aghion, B. and J. Morduch (2005). The Economics of Microfinance,
MIT Press.
Arnott, R. and J. E. Stiglitz (1991). "Moral Hazard and Non-Market Institutions:
Dysfunctional Crowding Out or Peer Monitoring." American Economic Review 81:
179-190.
Besley, T. J. and S. Coate (1995). "Group Lending, Repayment Incentives and Social
Collateral." Journal of Development Economics 46(1): 1-18.
Breza, E (2010) “Peer Effects and Loan Repayment: Evidence from the Krishna Default
Crisis”, mimeo, MIT.
13
Cassar, A., L. Crowley, B. Wydick (2007), “The Effect of Social Capital on Group
Loan Repayment: Evidence from Field Experiments,” Economic Journal, 117(517),
F85-F106.
Chowdhury, P. R. (2005). "Group-lending: Sequential financing, lender monitoring and
joint liability." Journal of Development Economics 77: 415-439.
Dowla, A. and D. Barua (2006). The Poor Always Pay Back: The Grameen II Story.
Kumarian Press, Inc, Bloomfield, CT
EDA Rural Systems, 2010, “Competition and the religion conundrum: The 2009
delinquency crisis in Southern Karnataka”. Gurgaon, India.
Feigenberg, B., E. Field and R. Pande (2010).
“Building Social Capital through
Microfinance.” Harvard University working paper.
Ghatak, M. (1999). "Group lending, local information and peer selection." Journal of
Development Economics 60(1): 27-50.
Ghatak, M. (2000). "Screening by the Company You Keep: Joint Liability Lending and
the Peer Selection Effect." Economic Journal 110(465).
Ghatak, M. and T. Guinnane (1999). "The Economics of Lending with Joint Liability: A
Review of Theory and Practice." Journal of Development Economics 60: 195-228.
Giné, X. and D. Karlan (2011) “Joint vs Individual Liability: Short and Long Term
Evidence from Philippine Microcredit Lending Groups”, mimeo.
Holmstrom, B. and P. Milgrom (1990). "Regulating Trade Among Agents." Journal of
Institutional and Theoretical Economics 146(1): 85-105.
Laffont, J. J. and P. Rey (2000). "Collusion and Group Lending with Moral Hazard."
Working Paper.
Matin, I. (1997). "Repayment performance of Grameen Bank borrowers: The 'unzipped'
state." Savings and Development 4.
Daley-Harris, S. (2009). "State of the Microcredit Summit Campaign Report 2009."
http://www.microcreditsummit.org/uploads/socrs/SOCR2009_English.pdf
Montgomery, R. (1996). "Disciplining or protecting the poor? Avoiding the social costs
of peer pressure in micro-credit schemes." Journal of International Development
8(2): 289-306.
Morduch, J. (1999). "The Microfinance Promise." Journal of Economic Literature
37(4): 1569-1614.
N'Guessan, T. and J. J. Laffont (2000). "Group Lending with Adverse Selection."
European Economic Review 44: 773-784.
14
Park, Albert and Ren (2011) “Microfinance with Chinese Characteristics” World
Development.
Rahman, Aminur (1999) “Microcredit Initiatives for Equitable and Sustainable
Development: Who Pays?” World Development 26 (12), December.
Rai, A. and T. Sjöström (2004). "Is Grameen Lending Efficient? Repayment Incentives
and Insurance in Village Economies." Review of Economic Studies 71(1): 217-234.
Rai, A. and T. Sjöström (2010). "Redisigning Microfinance", Willams College, mimeo.
Sadoulet, L. (2000). The Role of Mutual Insurance in Group Lending, ECARES/Free
Univ of Brussels.
Stiglitz, J. (1990). "Peer Monitoring and Credit Markets." World Bank Economic
Review 4(3): 351-366.
Varian, H. (1990). "Monitoring Agents With Other Agents." Journal of Institutional
and Theoretical Economics 146(1): 153-74.
Woolcock, M. (1999). "Learning from failures in microfinance: What unsuccessful
cases tell us about how group-based programs work." American Journal of
Economics and Sociology 58: 17-42.
15
Figures and Tables
Figure 1: Number of loans by date of maturity
1600
muslim
hindu
1400
1200
1000
800
600
400
200
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1
1
2
3
4
5
6
7
8
9
10
Figure 2: Average loan size by date of maturity
12000
muslim
hindu
10000
8000
6000
4000
2000
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1
1
2
16
3
4
5
6
7
8
9
10
Figure 3: Average interest rate by date of maturity
0.35
muslim
hindu
0.30
0.25
0.20
0.15
0.10
0.05
0.00
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1
1
2
3
4
5
6
7
8
9
10
Figure 4: Average number of installments by date of maturity
60
muslim
hindu
50
40
30
20
10
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1
1
2
17
3
4
5
6
7
8
9
10
Figure 5: Number of loans by date of disbursement (Kolar)
1200
muslim
hindu
1000
800
600
400
200
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1
1
2
3
4
5
6
7
8
9 10
Figure 6: Average loan size by date of disbursement (Kolar)
12000
muslim
hindu
10000
8000
6000
4000
2000
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1
1
18
2
3
4
5
6
7
8
9 10
Table 1: Comparison of baseline characteristics
All Borrowers
Borrowers in Mixed Centers
p-val
Hindus
Muslims
of ttest of
Borrowers with Multiple
Borrowers with Multiple
Loans
Loans in Mixed Centers
p-val
Hindus
Muslims
(1)-(2)
of ttest of
p-val
Hindus
Muslims
(4)-(5)
of ttest of
p-val
Hindus
Muslims
(7)-(8)
of ttest of
(7)-(8)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
Proportion of Muslims per center
0.25
0.82
-
0.42
0.63
-
0.23
0.92
-
0.40
0.70
-
Members per center
18.5
23.9
-
20.1
23.6
-
31.1
20.7
-
32.9
32.0
-
Number of centers
1,008
1,523
-
609
738
-
495
1023
-
279
277
-
Number of centers per borrower
1.11
1.16
0.000
1.16
1.22
0.000
2.17
2.25
0.000
2.19
2.30
0.000
Number of loans per borrower
1.55
1.64
0.016
1.61
1.76
0.011
3.31
3.35
0.467
3.35
3.51
0.017
15728
18134
-
6807
6759
-
1488
2300
-
926
1133
-
0.140
0.130
0.030
0.121
0.126
0.461
0.144
0.136
0.000
0.14
0.13
0.095
335
336
0.771
338
334
0.143
323
333
0.000
327
332
0.002
8579.7
8804.0
0.046
8650.1
8594.7
0.790
8002.5
8502.2
0.000
8021.8
8346.0
0.000
43.1
47.6
0.000
46
47
0.083
42
47
0.000
44
47
0.000
24,403
30,032
-
10,991
12,083
-
4931
7704
-
3101
3978
-
Panel A: Center Characteristics
Panel B: Borrower Characteristics
Number of borrowers
Panel C: Loan Characteristics
Interest rate
Loan duration (days)
Loan principal amount (Rs)
Number of installments
Number of loans
19
20
Table 2: Default at maturity (Hindu loans)
Hindus
Hindus
All
Hindus
in mixed
centers
Hindus
in mixed
with
centers
multiple
with
loans
multiple
loans
(1)
Post
(2)
(3)
(4)
0.101*** 0.192*** 0.154*** 0.224***
[0.032]
[0.071]
[0.034]
[0.054]
-0.002
-0.001
-0.002
-0.001
[0.002]
[0.002]
[0.001]
[0.001]
Pct Muslims in
Center
Pct Muslims x
Post
0.005*** 0.004*
0.005*** 0.004***
[0.002]
[0.002]
[0.001]
[0.001]
Mean Dep. Var
0.057
0.101
0.114
0.153
Observations
24,403
10,991
4,931
3,101
R-squared
0.702
0.717
0.571
0.593
Borrower FE included
21
Table 3: Default at maturity (Muslims loans)
Muslims
Muslims
Muslims
in mixed
All
in
with
centers
Muslims
mixed
multiple
with
centers
loans
multiple
loans
(1)
(2)
0.320**
0.275**
[0.149]
[0.139]
[0.120]
[0.118]
0.001
0.000
0.000
0.001
[0.001]
[0.001]
[0.001]
[0.001]
0.003*
0.004**
0.003**
0.003**
[0.002]
[0.002]
[0.001]
[0.001]
Mean Dep. Var
0.486
0.408
0.531
0.487
Observations
29,808
11,911
7,704
3,978
R-squared
0.804
0.772
0.659
0.619
Post
(3)
(4)
0.399*** 0.424***
Pct Muslims in
Center
Pct Muslims x
Post
Borrower FE included
22
Table 4: Balance outstanding at maturity (Hindu loans)
Hindus
Hindus
Hindus
in mixed
All
in
with
centers
Hindus
mixed
multiple
with
centers
loans
multiple
loans
(1)
Post
(2)
(3)
(4)
241.8**
517.3** 378.4*** 609.7***
[99.12]
[235.8]
[105.6]
[178.5]
-6.517
-3.148
-5.174
-2.557
[5.707]
[5.902]
[3.884]
[4.048]
Pct Muslims in
Center
Pct Muslims x Post
23.54*** 17.76** 21.16*** 16.61***
[6.354]
[7.216]
[4.888]
[5.423]
Mean Dep. Var
146.1
291.6
265.3
390
Observations
24,403
10,991
4,931
3,101
R-squared
0.671
0.673
0.447
0.447
Borrower FE included
23
Table 5: Balance outstanding at maturity (Muslim loans)
Hindus
All
Hindus
Hindus
in mixed
centers
Hindus
in mixed
with
centers
multiple
with
loans
multiple
loans
(1)
(2)
1,183
1,081
[752.3]
[696.0]
[799.2]
[780.2]
-9.761
-10.5
-3.447
-1.792
[8.034]
[7.517]
[5.871]
[5.623]
10.05
7.801
Post
(3)
(4)
2,133*** 2,240***
Pct Muslims in
Center
Pct Muslims x Post 20.73** 22.80***
[8.584]
[8.531]
[8.600]
[8.482]
2433
1849
2550
2157
Observations
29,808
11,911
7,704
3,978
R-squared
0.771
0.713
0.543
0.523
Mean Dep. Var
Borrower FE included
24
Table 6: Impact by density category
Loan Principal
Was loan in default
Amount
at maturity?
Balance
outstanding at
Pct. Loan due at
maturity
maturity
Hindus
Muslims
Hindus
Muslims
Hindus
Muslims
Hindus
Muslims
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
2,133***
2,328***
0.142***
0.574***
289.2**
2,804***
0.0528**
0.278***
(655.3)
(582.8)
(0.0389)
(0.0629)
(115.1)
(361.4)
(0.0245)
(0.0292)
1,712**
-819.3
-0.0928
-0.0343
-217.7
-851.8*
-0.0488
-0.0529
(734.0)
(832.6)
(0.0855)
(0.0655)
(207.4)
(455.1)
(0.0363)
(0.0326)
1,657*
-919.9
-0.0683
-0.120*
-183.9
-459.6
-0.02
-0.0504*
(921.1)
(677.0)
(0.0839)
(0.0626)
(276.1)
(360.0)
(0.0299)
(0.0273)
596.2
-303.5
-0.130*
-0.0359
-427.8*
-182.7
-0.0527*
-0.0253
(738.1)
(662.5)
(0.0680)
(0.0538)
(248.0)
(311.7)
(0.0311)
(0.0263)
165.1
99.61
0.144
0.0786
541.0*
201.9
0.0709
0.0348
(852.5)
(983.1)
(0.101)
(0.0917)
(314.0)
(575.1)
(0.0517)
(0.0418)
432.3
672.4
0.177*
0.0541
704.9**
-354.6
0.0479
0.000261
(961.3)
(709.4)
(0.0973)
(0.106)
(346.2)
(534.7)
(0.0397)
(0.0394)
263.1
250.9
0.285***
0.174**
1,226***
560.4
0.132***
0.0721**
(873.0)
(699.2)
(0.0770)
(0.0718)
(300.9)
(421.0)
(0.0433)
(0.0350)
Observations
3.101
3.978
3.101
3.978
3.101
3.978
3.101
3.978
R-squared
0.37
0.382
0.595
0.628
0.448
0.535
0.477
0.535
Mean Dep. Var.
8022
8346
0.153
0.487
390
2157
0.0468
0.226
Post
Quartile 2
Quartile 3
Quartile 4
Post x Quartile 2
Post x Quartile 3
Post x Quartile 4
We use the sample of borrowers with multiple loans in different centers. Standard errors in parentheses
(*** p<0.01, ** p<0.05, * p<0.1)
25